DB handles and radius.log errors

[Irina]

First, thanks to Ivan for help with Simultaneous and to Dennis with
indexing.

I am new to radius, please bear with me.  I will try to describe the problem
as much as I can.

I need to ask if anybody has experienced a problem with DB handles.  Here is
what we have experienced a couple of times.

Both NASes Cisco talk to 1 radius (there is another radius, but it was set
up on NASes as a secondary radius).

Many connections drop at the same time on 1 or both NASes and try to
reconnect.  Radius did not get a proper disconnection, therefore sees many
users connected with Stop = 0.  All Simultaneous Logins are set to 1, so
radius checks if users are connected on NAS.

Below is what I find in radius.log.

Mon Jun 11 14:48:28 2007 : Error: Discarding duplicate request from client
aleph:1645 - ID: 95 due to unfinished request 28856
Mon Jun 11 14:48:33 2007 : Error: Discarding duplicate request from client
aleph:1645 - ID: 95 due to unfinished request 28856
Mon Jun 11 14:48:33 2007 : Error: Check-TS: timeout waiting for checkrad

Mon Jun 11 14:48:48 2007 : Info: rlm_sql (sql): There are no DB handles to
use! skipped 0, tried to connect 0
Mon Jun 11 14:48:48 2007 : Info: rlm_sql (sql): There are no DB handles to
use! skipped 0, tried to connect 0

Also see logouts with no Login record
Mon Jun 11 14:48:52 2007 : Error: rlm_radutmp: Logout for NAS aleph port
1401, but no Login record
Mon Jun 11 14:48:52 2007 : Info: rlm_sql (sql): There are no DB handles to
use! skipped 0, tried to connect 0
Mon Jun 11 14:48:52 2007 : Error: rlm_radutmp: Logout for NAS aleph port
854, but no Login record

I also see few of the following for different usernames
Mon Jun 11 14:50:11 2007 : Error: rlm_sql (sql) in sql_accounting: stop
packet with zero session length. [user 'user at nas.net', nas '216.145.96.1']

Then
Mon Jun 11 14:50:16 2007 : Error: rlm_sql_getvpdata: database query error
Mon Jun 11 14:50:16 2007 : Error: rlm_sql (sql): SQL query error; rejecting
user


Also, radius.log has quite few lines (could be not related to only the
problem I described above, because I see the following lines in radius.log
constantly, with empty string instead of a username)

Wed Jun 13 09:39:26 2007 : Error: rlm_radutmp: Logout for NAS heh port 1099,
but no Login record
Wed Jun 13 09:39:26 2007 : Error: rlm_sql (sql) in sql_accounting: stop
packet with zero session length. [user '', nas '216.145.96.1']


I have increased num_sql_socks to 10.  Is it a reasonable number?  Or this
does not help with this problem and I need to look more into finding and
fixing it.

And/or NASes problem?


Any suggestion is welcome.  Thank you for your help in advance.

Irina
====================